Method for predicting concentration of dissolved gas in transformer oil based on EVO-VMD-LSSVM

Abstract

The invention discloses a method for predicting the concentration of dissolved gas in transformer oil based on an EVO-VMD-LSSVM. An energy valley optimization algorithm EVO, variable mode decomposition VMD and a least square support vector machine LSSVM are combined and used. The method comprises the following steps: decomposing an original dissolved gas concentration signal into a plurality of intrinsic mode functions IMFs with different frequency characteristics through a VMD method, and extracting multi-scale characteristics of the signal; a plurality of factors such as transformer load, operating oil temperature, environment temperature and operating time are introduced into the model, an LSSVM prediction model is established for IMF in combination with characteristic parameter correlation, LSSVM model parameters are optimized by using an EVO algorithm, and prediction precision is improved; reconstructing the prediction result of each IMF to obtain a final dissolved gas concentration prediction value; the prediction performance is obviously better than that of a traditional method, and evaluation indexes such as mean square error (MSE), mean absolute error (MAE) and decision coefficient (R) show that the method has high prediction precision and robustness.

Description

Technical Field

[0001] The invention relates to the technical field of power system equipment status monitoring, and in particular to a method for predicting dissolved gas concentration in transformer oil based on EVO-VMD-LSSVM. Background Art

[0002] In the power system, the transformer, as one of the key equipment, undertakes the important tasks of voltage conversion, power distribution and transmission. Its operating status directly affects the safety and stability of the entire power grid. In order to ensure the reliable operation of the transformer, it is very important to monitor its status in a timely and accurate manner. Dissolved Gas Analysis (DGA) in transformer oil is one of the most widely used transformer status monitoring technologies. This technology analyzes the gases (such as hydrogen) dissolved in the transformer oil. , Methane , Ethylene , acetylene , ethane , Carbon monoxide ,carbon dioxide ) can be analyzed to infer the type of fault inside the transformer and its severity, providing an important basis for the maintenance and overhaul of the transformer.

[0003] However, the traditional DGA method mainly relies on empirical formulas and expert judgment, which has certain subjectivity and limitations. Since the concentration of dissolved gas in transformer oil is affected by many factors, its time series data contains many components of different scales and trends, and has strong nonlinear and non-stationary characteristics, which makes it difficult for the traditional DGA method to meet the high requirements of modern power systems for transformer fault prediction in terms of prediction accuracy and real-time performance.

[0004] In order to overcome the shortcomings of the traditional DGA method, data-driven prediction methods have gradually attracted the attention of researchers in recent years. Among them, the least squares support vector machine (LSSVM), as an efficient machine learning algorithm, is widely used in various prediction models due to its advantages in dealing with small samples and nonlinear problems. However, the direct application of LSSVM to predict the concentration of dissolved gas in transformer oil still faces many challenges. On the one hand, the dissolved gas concentration signal usually has strong non-stationary and multi-scale characteristics, which is difficult to be effectively modeled by a single machine learning model; on the other hand, the performance of the LSSVM model depends on the selection of parameters. How to optimize the model parameters to improve the prediction accuracy is an urgent problem to be solved.

[0005] For example, CN117894389A discloses a method for predicting dissolved gas concentration data in transformer oil based on SSA optimization of VMD and LSTM, which includes the following steps: S1. Optimizing the number of decomposition modes K and penalty parameter ɑ of the VMD (Variational Mode Decomposition) model using the SSA (Sparrow Search Algorithm) algorithm; S2. Decomposing the monitored dissolved gas concentration data in transformer oil into multiple subsequences using the optimized VMD model; S3. Optimizing the number of neurons in the hidden layer, the number of iterations and the learning rate parameters of the LSTM (Long Short-Term Memory) model using the SSA algorithm; S4. Predicting each subsequence using the optimized LSTM model, superimposing and evaluating the numerical prediction results; S5. Constructing a set of prediction results at each moment, calculating the adaptive optimal window width of each set, and fitting the prediction interval of dissolved gas concentration in transformer oil of KDE according to the optimal window width and Gaussian kernel function. This method optimizes the variational mode decomposition parameters based on the sparrow search algorithm to perform mode decomposition on the monitoring data, and uses SSA to optimize the long short-term memory network model, which can improve the prediction accuracy; however, the structure is relatively complex, has high requirements for computing resources, and has a slow convergence speed. Especially when dealing with complex high-dimensional problems, the optimization efficiency may be insufficient. When the data diversity is insufficient or the sample size is small, the generalization ability may be limited, and overfitting is prone to occur. At the same time, the current prediction model or method of dissolved gas in transformer oil does not consider the possible influence of other external factors such as transformer load, oil temperature, ambient temperature, and operating time on the concentration of dissolved gas in oil, resulting in insufficient prediction stability and anti-interference ability.

[0006] In summary, although the existing prediction method of dissolved gas concentration in transformer oil based on LSSVM has improved the prediction performance to a certain extent, it still has problems such as insufficient prediction accuracy, difficulty in optimizing model parameters, and poor resistance to external interference. Therefore, there is an urgent need for a new method that can more effectively extract the multi-scale features of dissolved gas concentration signals, integrate internal and external factors to construct and optimize LSSVM model parameters, and thus improve prediction accuracy and real-time performance. Summary of the invention

[0007] The technical problem to be solved by the present invention is to provide a method for predicting the dissolved gas concentration in transformer oil based on EVO-VMD-LSSVM, so as to solve the technical problems existing in the prediction of the dissolved gas concentration in transformer oil in the existing power system, especially the traditional DGA method which relies on empirical formulas and expert judgment, has subjectivity and limitations, and the challenges faced by the prediction model based on LSSVM when processing dissolved gas concentration signals with non-stationary and multi-scale characteristics.

[0008] In order to solve the above technical problems, the technical solution adopted by the present invention is: a method for predicting dissolved gas concentration in transformer oil based on EVO-VMD-LSSVM, comprising the following steps: Step 1: Data preprocessing, including denoising and normalization processing, is performed on the data of dissolved gas concentration in transformer oil and parameters such as transformer load, operating oil temperature, ambient temperature, and operating time. Step 2: VMD decomposition: input the preprocessed dissolved gas concentration signal into the VMD algorithm and decompose it into several intrinsic mode functions (IMFs) to extract the multi-scale features of the signal; Step 3: LSSVM modeling. For each IMF, select the gas concentration parameter IMF sequence with a correlation greater than 0.6 and external parameters such as transformer load, oil temperature, ambient temperature, and operating time as the prediction model input, establish the LSSVM regression model, and set the main parameters of the LSSVM model including the penalty parameter C and the kernel function parameter ; Step 4: EVO (Energy valley optimizer) optimization is a meta-heuristic algorithm for solving global optimization problems. It simulates the process of unstable particles changing the ratio of neutrons and protons that make up the particles through α, β, γ decay and random movement, so that unstable particles move to the stable belt or energy valley to increase their stability level. The EVO algorithm is used to optimize the parameters of the LSSVM model to improve the prediction accuracy. Step 5: Reconstruct the prediction results. Reconstruct the prediction results of each IMF to obtain the final predicted value of dissolved gas concentration. Step 6: Model verification, verify the prediction performance of the model through actual data.

[0009] In a preferred solution, the data preprocessing in Step 1 also includes filling the missing data using linear interpolation.

[0010] In a preferred solution, the VMD decomposition in Step 2 specifically includes the following steps: Step 2.1: Initialization, setting the initial frequency center and bandwidth parameters; Step 2.2: Iterative optimization, by iteratively updating the frequency center and bandwidth parameters, the bandwidth of each mode is minimized; Step 2.3: Modal reconstruction: bandpass filter the signal according to the optimized frequency center and bandwidth parameters to obtain each IMF.

[0011] In a preferred solution, the EVO optimization in Step 4 specifically includes the following steps: Step 4.1: Initial population generation: randomly generate a set of initial solutions in the parameter space to form the initial population; initialize a population containing N candidate solutions, each candidate solution is randomly generated in the search space: (1) In the formula, For the Particle No. dimensional space position, , Respectively Particle No. The upper and lower limits of the dimension variable; is a random number in the range [0,1]; is the particle size in the search space; is the dimension of decision variables.

[0012] Step 4.2: Calculate the fitness value, i.e., the neutron enrichment level of the particle, and the enrichment boundary of the particle group , calculate the stability level of the particle : (2) (3) In the formula, is the enrichment limit of particles in the search space; For the The neutron enrichment level of each particle; and Represent the fitness function values ​​of the particles with the best and worst stability respectively.

[0013] Step 4.3: and Decay position update, if the The stability level of each particle , then occurs and decay, Position of new decay particles: (4) In the formula, In the search space The positions of new particles generated by decay; is the first The current position of each particle; is the particle position with the best stability; For the decision variables.

[0014] Position of new decay particles: (5) In the formula, In the search space The positions of new particles generated by decay; For the The particle positions around a particle.

[0015] Step 4.4: Decay position update, if the The stability level of each particle , then occurs decay, Position of new decay particles: (6) (7) In the formula, In the search space The positions of new particles generated by decay; , are two random numbers in the range of [0, 1]; is the center position of the particle.

[0016] use The decaying particle undergoes another position update process: (8) In the formula, is the position of another new particle generated by decay in the search space; , is in the range [0, 1] and is different from , Two random numbers.

[0017] Step 4.5: Randomly move the position to update, if the neutron enrichment level is lower than the enrichment limit , the particle performs electron capture or positron emission to move to a stable position, and the position is updated as: (9) In the formula, Generate new particle positions randomly in the search space; is a random number in the range [0, 1].

[0018] Step 4.6: Determine whether the termination condition is met. If the algorithm reaches the maximum number of iterations, exit the loop and output Otherwise, return to step 4.2 to continue.

[0019] In the preferred solution, the model validation in Step 5 uses mean square error MSE, mean absolute error MAE and determination coefficient R² as evaluation indicators.

[0020] In the preferred solution, the mean square error (MSE) is a measure of the average square error between the predicted value and the actual value, and its expression is: (10) In the formula, is the length of the data sequence; is the serial number position corresponding to each moment sequence; For the The actual value of the parameter; For the The predicted value of a parameter.

[0021] In the preferred solution, the mean absolute error (MAE) is the average absolute error between the predicted value and the actual value, and its expression is: (11) In the preferred solution, the coefficient of determination (R²) is a measure of the degree of fit of the model to the data, and its expression is: (12) In the formula, is the average of the measured values ​​of the parameter time series.

[0022] The method for predicting dissolved gas concentration in transformer oil based on EVO-VMD-LSSVM provided by the present invention has the following beneficial effects: 1. The present invention solves the technical problems existing in the prediction of dissolved gas concentration in transformer oil in the existing power system, especially the problem that the traditional DGA method mainly relies on empirical formulas and expert judgment, has subjectivity and limitations, and is difficult to meet the high-precision and real-time requirements of modern power systems for transformer fault prediction; 2. The present invention proposes to construct a prediction model by combining multi-factor analysis and feature fusion, which not only considers the time series characteristics of the concentration of dissolved gas in oil, but also introduces external factors such as transformer load, operating oil temperature, ambient temperature and operating time to conduct multivariate correlation analysis. This multi-factor fusion processing method enables the model to more comprehensively reflect the actual working conditions and improves the adaptability and accuracy of the prediction; 3. The present invention decomposes the original dissolved gas concentration signal into multiple IMFs through VMD decomposition, effectively extracts the multi-scale characteristics of the signal, and solves the problem that the dissolved gas concentration signal usually has strong non-stationary and multi-scale characteristics, which makes it difficult to effectively model it through a single machine learning model and difficult to process nonlinear and non-stationary data; 4. The present invention uses the EVO algorithm to optimize the parameters of the LSSVM model, significantly improving the prediction accuracy and robustness of the model, making the prediction results more accurate and reliable; and enhancing the ability to handle nonlinear and high-dimensional complex relationships. LSSVM shows good generalization performance in small sample conditions, making up for the shortcomings of the neural network-based method in terms of long training time and high risk of overfitting; 5. The present invention organically combines VMD, LSSVM and EVO to form an efficient method for predicting dissolved gas concentration, which overcomes the technical bottlenecks of model parameter optimization and external factors affecting the prediction accuracy when directly applying LSSVM to predict the dissolved gas concentration in transformer oil in the prior art, and can effectively support the real-time monitoring and fault prediction of the dissolved gas concentration in transformer oil; 6. The present invention extracts multi-scale features through VMD decomposition and optimizes LSSVM model parameters through EVO, which significantly improves the prediction accuracy of dissolved gas concentration; 7. The method of the present invention can effectively extract the multi-scale characteristics of the dissolved gas concentration signal. Through VMD decomposition, the original signal is decomposed into several IMFs, each of which represents the component of the signal in different frequency ranges, thereby accurately reflecting the nonlinear and non-stationary characteristics of the signal; 8. The method of the present invention realizes real-time monitoring of dissolved gas concentration in transformer oil and fault prediction, providing a strong guarantee for the safe and stable operation of the power system; 9. The method of the present invention provides a new and effective means for transformer status monitoring and fault prediction, and has broad application prospects and practical value; 10. The present invention also has high flexibility and scalability, and can be adaptively adjusted according to the needs of different application scenarios, further enhancing its competitiveness in the field of prediction of dissolved gas concentration in transformer oil. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] The present invention will be further described below in conjunction with the accompanying drawings and implementation examples: Figure 1 It is a schematic diagram of the technical process of the present invention. DETAILED DESCRIPTION

[0024] The technical solution of the present invention is further described below in conjunction with the accompanying drawings and embodiments: Example 1 like Figure 1 As shown, the prediction method of dissolved gas concentration in transformer oil based on EVO-VMD-LSSVM includes the following steps: Step 1: Data preprocessing, including denoising and normalization processing, is performed on the data of dissolved gas concentration in transformer oil and parameters such as transformer load, operating oil temperature, ambient temperature, and operating time. Step 2: VMD decomposition: input the preprocessed dissolved gas concentration signal into the VMD algorithm and decompose it into several intrinsic mode functions (IMFs) to extract the multi-scale characteristics of the signal. Step 3: LSSVM modeling. For each IMF, select the gas concentration parameter IMF sequence with a correlation greater than 0.6 and external parameters such as transformer load, oil temperature, ambient temperature, and operating time as the prediction model input, establish the LSSVM regression model, and set the main parameters of the LSSVM model including the penalty parameter C and the kernel function parameter ; Step 4: EVO (Energy valley optimizer) optimization, using the EVO algorithm to optimize the parameters of the LSSVM model to improve the prediction accuracy; Step 5: Reconstruct the prediction results, reconstruct the prediction results of each IMF to obtain the final predicted value of dissolved gas concentration; Step 6: Model verification, verify the prediction performance of the model through actual data.

[0025] In this embodiment, the data preprocessing in Step 1 also includes filling the missing data using a linear interpolation method.

[0026] Furthermore, the VMD decomposition in Step 2 specifically includes the following steps: Step 2.1: Initialization, setting the initial frequency center and bandwidth parameters; Step 2.2: Iterative optimization, by iteratively updating the frequency center and bandwidth parameters, the bandwidth of each mode is minimized; Step 2.3: Modal reconstruction: bandpass filter the signal according to the optimized frequency center and bandwidth parameters to obtain each IMF.

[0027] Furthermore, the EVO optimization in Step 4 specifically includes the following steps: Step 4.1: Initial population generation: randomly generate a set of initial solutions in the parameter space to form the initial population; Step 4.2: Energy calculation, calculate the energy value of each individual according to the prediction error; Step 4.3: Individual update, generating new individuals through selection, crossover and mutation operations; Step 4.4: Energy valley search, through local search, further reduce the energy value of the individual; Step 4.5: Determine whether the termination condition is met. If so, terminate the algorithm.

[0028] Furthermore, the model validation in Step 5 uses mean square error MSE, mean absolute error MAE and determination coefficient R² as evaluation indicators.

[0029] Furthermore, the mean square error (MSE) is a measure of the average square error between the predicted value and the actual value, and its expression is: (10) In the formula, is the length of the data sequence; is the serial number position corresponding to each moment sequence; For the The actual value of the parameter; For the The predicted value of a parameter.

[0030] Furthermore, the mean absolute error (MAE) is the average absolute error between the predicted value and the actual value, and its expression is: (11) Furthermore, the coefficient of determination (R²) is a measure of how well the model fits the data, and its expression is: (12) In the formula, is the average of the measured values ​​of the parameter time series.

[0031] Example 2 In another preferred embodiment, on the basis of Example 1, the rest of the steps are the same as those in Example 1. This embodiment further refines and specifies the EVO optimization step in Step 4: Step 4.1: Initial population generation: randomly generate a set of initial solutions in the parameter space to form the initial population; initialize a population containing N candidate solutions, each candidate solution is randomly generated in the search space: (1) In the formula, For the Particle No. dimensional space position, , Respectively Particle No. The upper and lower limits of the dimension variable; is a random number in the range [0,1]; is the particle size in the search space; is the dimension of decision variables.

[0032] Step 4.2: Calculate the fitness value, i.e., the neutron enrichment level of the particle, and the enrichment boundary of the particle group , calculate the stability level of the particle : (2) (3) In the formula, is the enrichment limit of particles in the search space; For the The neutron enrichment level of each particle; and Represent the fitness function values ​​of the particles with the best and worst stability respectively.

[0033] Step 4.3: and The decay position is updated. If The stability level of each particle , then occurs and decay, Position of new decay particles: (4) In the formula, In the search space The positions of new particles generated by decay; is the first The current position of each particle; is the particle position with the best stability; For the decision variables.

[0034] Position of new decay particles: (5) In the formula, In the search space The positions of new particles generated by decay; For the The particle positions around a particle.

[0035] Step 4.4: Decay position update, if the The stability level of each particle , then occurs decay, Position of new decay particles: (6) (7) In the formula, In the search space The positions of new particles generated by decay; , are two random numbers in the range of [0, 1]; is the center position of the particle.

[0036] use The decaying particle undergoes another position update process: (8) In the formula, is the position of another new particle generated by decay in the search space; , is in the range [0, 1] and is different from , Two random numbers.

[0037] Step 4.5: Randomly move the position to update, if the neutron enrichment level is lower than the enrichment limit , the particle performs electron capture or positron emission to move to a stable position, and the position is updated as: (9) In the formula, Generate new particle positions randomly in the search space; is a random number in the range [0, 1].

[0038] Step 4.6: Determine whether the termination condition is met. If the algorithm reaches the maximum number of iterations, exit the loop and output Otherwise, return to step 4.2 to continue.

[0039] Example 3 In another preferred embodiment, based on embodiments 1 and 2, refer to Figure 1 This embodiment describes in detail the method for predicting dissolved gas concentration in transformer oil based on EVO-VMD-LSSVM. The specific steps are as follows: 1. Data preprocessing First, the data of dissolved gas concentration in transformer oil is preprocessed; since there may be noise and outliers in the original data, denoising is required. Then, the missing data is filled. This embodiment adopts linear interpolation method, assuming that the data changes smoothly near the missing point, and estimates the missing value by linear interpolation; finally, in order to eliminate the dimensional differences between different indicators and improve the convergence speed and prediction accuracy of model training, the data is normalized and the minimum-maximum normalization method is used to map the data to the [0, 1] interval.

[0040] 2. VMD decomposition The pre-processed dissolved gas concentration signal is input into the VMD algorithm for decomposition; the VMD algorithm is a variable mode decomposition method that can better achieve the stabilization of nonlinear and non-stationary data; through VMD decomposition, the original signal can be decomposed into several intrinsic mode functions (IMFs) with different frequency characteristics. The specific decomposition steps are as follows: 2.1 Initialization: set the initial frequency center and bandwidth parameters; 2.2 Iterative optimization: The bandwidth of each mode is minimized by iteratively updating the frequency center and bandwidth parameters; 2.3 Modal reconstruction: According to the optimized frequency center and bandwidth parameters, the signal is bandpass filtered to obtain the various IMFs.

[0041] In this embodiment, the VMD algorithm is selected through EVO optimization, and the number of decomposition modes and the penalty coefficient are determined to further optimize the decomposition effect.

[0042] 3. LSSVM modeling According to the characteristics of each IMF, the gas concentration parameter IMF sequence with a correlation greater than 0.6 and external parameters such as transformer load, oil temperature, ambient temperature, and operating time are selected as the prediction model input to establish the LSSVM regression model; the LSSVM model is an efficient machine learning algorithm suitable for processing small samples and nonlinear problems; the main parameters of the LSSVM model include the penalty parameter C and the kernel function parameter ; The model is trained by historical data to determine the model parameters; In this embodiment, the radial basis function (RBF) is used as the kernel function.

[0043] 4. EVO Optimization The EVO algorithm is used to optimize the parameters of the LSSVM model to improve the prediction accuracy; the EVO algorithm is an energy valley optimization algorithm that can find the global optimal solution. The specific steps are as follows: 4.1 Initial population generation: A set of initial solutions are randomly generated in the parameter space to form the initial population; 4.2 Energy calculation: Calculate the energy value of each individual based on the prediction error. The smaller the energy value, the higher the fitness of the individual. 4.3 Individual position update: Generate new individuals through selection, crossover and mutation operations; 4.4 Energy Valley Search: Further reduce the energy value of individuals through local search; 4.5 Termination condition judgment: judge whether the termination condition is met, if so, terminate the algorithm, otherwise return to the energy calculation step; 4.6 Through the EVO algorithm, the optimal penalty parameters and kernel function parameters of the LSSVM model can be found , which significantly improves the prediction accuracy of the model.

[0044] 5. Reconstruction of prediction results The prediction results of each IMF are reconstructed to obtain the final predicted value of dissolved gas concentration; specifically, the predicted value of each IMF is added together to obtain the reconstructed predicted value.

[0045] 6. Model Validation The prediction performance of the model was verified by actual data. Evaluation indicators such as mean square error (MSE), mean absolute error (MAE) and determination coefficient (R²) were used to comprehensively evaluate the prediction accuracy and fitting ability of the model; the experimental results show that the method proposed in this paper is significantly better than the traditional method in terms of prediction accuracy.

[0046] Example 3 In another preferred embodiment, based on embodiments 1 and 2, refer to Figure 1 In order to demonstrate the diversity and practicality of the prediction method of the present invention, this embodiment provides another specific application scenario.

[0047] Different from Example 2, this example adopts different parameter settings and optimization strategies in the VMD decomposition and LSSVM modeling processes.

[0048] 1. Data preprocessing Similar to Example 2, the data is denoised, missing values ​​are filled, and normalized.

[0049] 2. VMD decomposition The same VMD algorithm as in Example 2 is used for decomposition, but the selection of the number of decomposition modes and the penalty coefficient is different; this example determines the optimal number of decomposition modes and the penalty coefficient through multiple experiments to further improve the decomposition effect.

[0050] 3. LSSVM modeling When establishing the LSSVM regression model, different penalty parameters C and kernel function parameters are used At the same time, this embodiment also tries different kernel function types, such as polynomial kernel function and Sigmoid kernel function, to find the kernel function type that best suits the present invention.

[0051] 4. EVO Optimization Similar to Example 2, the EVO algorithm is used to optimize the parameters of the LSSVM model; however, different strategies are used in the individual update and energy valley search process to further improve the optimization effect.

[0052] 5. Reconstruction of prediction results The prediction results of each IMF are reconstructed to obtain the final predicted value of dissolved gas concentration.

[0053] 6. Model Validation Similar to Example 2, the prediction performance of the model is verified by actual data, and the same evaluation indicators are used to comprehensively evaluate the prediction accuracy and fitting ability of the model; the experimental results show that after adopting the parameter setting and optimization strategy of this embodiment, the prediction accuracy and fitting ability of the model are significantly improved, and the stability and reliability of the prediction results are also stronger, which further verifies the diversity and practicality of the prediction method of the present invention.

[0054] In the preferred solution, the data preprocessing in Step 1 also includes filling the missing data with linear interpolation. The above settings can effectively reduce the analysis errors caused by missing data and improve the accuracy of the model. At the same time, the interquartile range method is used to clean the outliers to ensure the quality of the input data and lay a solid foundation for subsequent steps.

[0055] In the preferred solution, the model verification in Step 5 uses mean square error MSE, mean absolute error MAE and determination coefficient R² as evaluation indicators; the above settings are intended to comprehensively measure the prediction performance of the model, among which MSE evaluates the sum of squares of the deviations between the predicted value and the true value, MAE directly reflects the average level of prediction error, and R² reveals the model's ability to explain variable fluctuations.

[0056] In summary, the method for predicting dissolved gas concentration in transformer oil based on EVO-VMD-LSSVM proposed in the present invention solves the technical problems existing in the prediction of dissolved gas concentration in transformer oil in the existing power system, especially the problem that the traditional DGA method mainly relies on empirical formulas and expert judgment, has subjectivity and limitations, and is difficult to meet the high-precision and real-time requirements of transformer fault prediction in modern power systems; the present invention innovatively introduces the variable mode decomposition (VMD) method to decompose the original signal into several intrinsic mode functions (IMFs) with different frequency characteristics, thereby realizing the accurate extraction of multi-scale features of the signal. This method not only improves the accuracy of signal processing, It also provides richer feature information for subsequent prediction models; the present invention aims at the difficulty of least squares support vector machine (LSSVM) model parameter optimization, and this scheme adopts the energy valley optimization (EVO) algorithm; as an emerging optimization technology, the EVO algorithm has significant advantages in parameter search and optimization; by optimizing the LSSVM model parameters through the EVO algorithm, the prediction accuracy and generalization ability of the model can be significantly improved, thereby achieving a more accurate prediction of the concentration of dissolved gas in transformer oil; the prediction model is constructed by combining multi-factor analysis and feature fusion, which not only considers the time series characteristics of the concentration of dissolved gas in oil, but also introduces external factors such as transformer load, operating oil temperature, ambient temperature and operating time. factors, and conduct multivariate correlation analysis. This multi-factor fusion processing method enables the model to more comprehensively reflect the actual working conditions and improves the adaptability and accuracy of the prediction. The present invention organically combines the three methods of VMD, LSSVM and EVO to construct a new comprehensive prediction model. This model not only integrates the advantages of each method, but also realizes real-time monitoring and fault prediction of dissolved gas concentration in transformer oil through mutual synergy. This comprehensive prediction model has high novelty in theory and practice. By introducing the VMD method and EVO algorithm, the present invention realizes technical innovation in signal processing, model construction and parameter optimization. These innovations not only improve the accuracy of the prediction, but also The invention not only improves the accuracy and efficiency of transformer status monitoring and fault prediction, but also provides a new technical path for transformer status monitoring and fault prediction. The invention organically combines VMD, LSSVM and EVO to form a new comprehensive prediction method, which not only overcomes the limitations of traditional methods, but also realizes a more accurate prediction of the concentration of dissolved gas in transformer oil through innovative combination. The invention has broad application prospects in the field of transformer status monitoring and fault prediction. By real-time monitoring of the concentration change of dissolved gas in transformer oil, potential fault hazards can be discovered in time, providing strong guarantee for the safe and stable operation of the power system. At the same time, the scheme can also provide reference and reference for status monitoring and fault prediction in other fields.The present invention proposes an efficient and accurate prediction scheme for the concentration of dissolved gas in transformer oil, which not only promotes the development of transformer condition monitoring technology, but also provides important technical support for the construction and operation and maintenance of smart grids, and has important scientific significance and practical application value. ;

Claims

1. The method for predicting dissolved gas concentration in transformer oil based on EVO-VMD-LSSVM is characterized by: The prediction method comprises the following steps: Step 1: Data preprocessing, including denoising and normalization processing, is performed on the data of dissolved gas concentration in transformer oil and parameters such as transformer load, operating oil temperature, ambient temperature, and operating time. Step 2: VMD decomposition: input the preprocessed dissolved gas concentration signal into the VMD algorithm and decompose it into several intrinsic mode functions (IMFs) to extract the multi-scale characteristics of the signal. Step 3: LSSVM modeling: For each IMF, an LSSVM regression model is established. The main parameters of the LSSVM model include the penalty parameter C and the kernel function parameter ; Step 4: EVO optimization, using the EVO algorithm to optimize the parameters of the LSSVM model to improve the prediction accuracy; Step 5: Reconstruct the prediction results, reconstruct the prediction results of each IMF to obtain the final predicted value of dissolved gas concentration; Step 6: Model verification, verify the prediction performance of the model through actual data.

2. The method for predicting dissolved gas concentration in transformer oil according to claim 1, wherein: The data preprocessing in Step 1 also includes filling the missing data using linear interpolation.

3. The method for predicting dissolved gas concentration in transformer oil according to claim 1, wherein: The VMD decomposition in Step 2 is as follows: The following steps are involved: Step 2.1: Initialization, setting the initial frequency center and bandwidth parameters; Step 2.2: Iterative optimization, by iteratively updating the frequency center and bandwidth parameters, the bandwidth of each mode is minimized; Step 2.3: Modal reconstruction: bandpass filter the signal according to the optimized frequency center and bandwidth parameters to obtain each IMF.

4. The method for predicting dissolved gas concentration in transformer oil according to claim 1, wherein: The LSSVM modeling in Step 3 is to select the gas concentration parameter IMF sequence with a correlation greater than 0.6 and external parameters such as transformer load, oil temperature, ambient temperature, and operating time as the prediction model input for each IMF to establish the LSSVM regression model.

5. The method for predicting dissolved gas concentration in transformer oil according to claim 1, wherein: The EVO optimization in Step 4 specifically includes the following steps: Step 4.1: Initial population generation: randomly generate a set of initial solutions in the parameter space to form the initial population; initialize a population containing N candidate solutions, each candidate solution is randomly generated in the search space: （1）； In the formula, For the Particle No. dimensional space position, , Respectively Particle No. The upper and lower limits of the dimension variable; is a random number in the range [0,1]; is the particle size in the search space; is the dimension of decision variables; Step 4.2: Calculate the fitness value, i.e., the neutron enrichment level of the particle, and the enrichment boundary of the particle group , calculate the stability level of the particle : （2）； （3）； In the formula, is the enrichment limit of particles in the search space; For the The neutron enrichment level of each particle; and Represent the fitness function values ​​of the particles with the best and worst stability respectively; Step 4.3: and Decay position update, if the The stability level of each particle , then occurs and decay, Position of new decay particles: （4）； In the formula, In the search space The positions of new particles generated by decay; is the first The current position of each particle; is the particle position with the best stability; For the decision variables; Position of new decay particles: （5）； In the formula, In the search space The positions of new particles generated by decay; For the The positions of particles around a particle; Step 4.4: Decay position update, if the The stability level of each particle , then occurs decay, Position of new decay particles: （6）； （7）； In the formula, In the search space The positions of new particles generated by decay; , are two random numbers in the range of [0, 1]; is the center position of the particle; use The decaying particle undergoes another position update process: （8）； In the formula, is the position of another new particle generated by decay in the search space; , is in the range [0, 1] and is different from , Two random numbers; Step 4.5: Randomly move the position to update, if the neutron enrichment level is lower than the enrichment limit , the particle performs electron capture or positron emission to move to a stable position, and the position is updated as: （9）； In the formula, Generate new particle positions randomly in the search space; is a random number in the range of [0, 1]; Step 4.6: Determine whether the termination condition is met. If the algorithm reaches the maximum number of iterations, exit the loop and output Otherwise, return to step 4.2 to continue.

6. The method for predicting dissolved gas concentration in transformer oil according to claim 1, characterized in that: The model validation in Step 5 uses mean square error MSE, mean absolute error MAE and determination coefficient R² as evaluation indicators.

7. The method for predicting dissolved gas concentration in transformer oil according to claim 6, characterized in that: The mean square error MSE is a measure of the average square error between the predicted value and the actual value, and its expression is: （10）； In the formula, is the length of the data sequence; is the serial number position corresponding to each moment sequence; For the The actual value of the parameter; For the The predicted value of a parameter.

8. The method for predicting dissolved gas concentration in transformer oil according to claim 6, characterized in that: The mean absolute error MAE is a measure of the mean absolute error between the predicted value and the actual value, and its expression is: （11）。 9. The method for predicting dissolved gas concentration in transformer oil according to claim 6, characterized in that: The coefficient of determination R² is a measure of how well the model fits the data, and its expression is: （12）； In the formula, is the average of the measured values ​​of the parameter time series.

Images